Calibration of Serial Manipulators: Theory and Applications

The kinematic calibration is a procedure to improve the manipulator accuracy without mechanical means by acting on the manipulator controller. Although manipulators are composed by accurate mechanical components, the precision of their motion is affected by many sources of error (Mooring et al, 1991). The final position accuracy is mainly influenced by: kinematic inaccuracy (due to manufacturing and assembly errors in both actuated and passive joints), load deformation (due to external forces including gravity) and thermal deformation (Reinhar et al, 2004). This is true for serial (Mooring et al, 1991) as for parallel (Wildenberg, 2000) manipulators as well. Each of these factors should be addressed with an appropriate compensation or calibration methodology. This work deals with kinematic inaccuracy, related to robot geometry, assembly and joint motion. One possibility to compensate for geometrical errors is to perform a kinematic calibration. The robot is requested to reach some desired poses and the reached actual poses are measured. Then, the exact robot geometry is estimated analyzing the difference between the desired and the reached poses. This procedure requires a parametric identification of the manipulator which consists in the formulation of a geometrical model of the robot in which each source of error is represented by a parameter. The parameter set includes link lengths, joint axes inclination and joint coordinate offsets. The calibration consists in identifying the actual values of all these parameters. Once this operation is performed, it is possible to predict the pose error for any robot configuration and so it is possible to compensate for them by suitable joint motions. The aim of this work is to address all the steps of the procedure which includes: 1. The development of a suitable kinematic model of a general serial manipulator; 2. One example of collection of experimental data; 3. The estimation of the numerical value of the manipulator parameters; 4. The error compensation. For each phase different alternatives are described and critically compared. A relevant part of the chapter summarises, compares and extends the existing techniques used to generate a suitable parametric kinematic model for a general serial manipulator. Rules for automatic generation of models with different characteristics are developed. These results will be used in a next chapter for the development of a parametric kinematic model for a general parallel manipulator (PKM).